# Hugging Face Transformers v3.4.0 — Release & Upgrade Intelligence
- **Product:** Hugging Face Transformers (https://huggingface.co)
- **Version:** v3.4.0
- **Released:** Oct 20, 2020
- **Impact Score:** 7/10
- **Breaking Changes:** Source flagged
- **Canonical Analysis:** https://changelogs.us/transformers/updates/v3.4.0/

## Executive Summary (TL;DR)
Hugging Face Transformers v3.4.0. Source highlight: ProphetNet, Blenderbot, SqueezeBERT, DeBERTa

## Breaking Changes & Deprecations
- - Adding Fast tokenizers for SentencePiece based tokenizers - Breaking: remove Transfo-XL fast tokenizer #7141 (@thomwolf)

## New Features
- Add DeBERTa model #5929 (@BigBird01)
- Add specific notebook ProgressCalback #7793 (@sgugger)
- Add predict step accumulation #7767 (@sgugger)
- [s2s] add config params like Dropout in Seq2SeqTrainingArguments #7532 (@patil-suraj)

## Improvements
- Remove config assumption in Trainer #7464 (@sgugger)
- Clean the Trainer state #7490 (@sgugger)
- Small QOL improvements to TrainingArguments #7475 (@sgugger)
- Allow nested tensors in predicted logits #7542 (@sgugger)
- Trainer callbacks #7596 (@sgugger)
- Don't use `store_xxx` on optional bools #7786 (@sgugger)

## Bug Fixes
- Fix squeezebert docs #7587 (@LysandreJik)
- Fix DeBERTa integration tests #7729 (@LysandreJik)
- Small fixes to NotebookProgressCallback #7813 (@sgugger)
- Distributed Trainer: 2 little fixes #7461 (@sshleifer)
- [s2strainer] fix eval dataset loading #7477 (@patil-suraj)
- [s2s] Fix t5 warning for distributed eval #7487 (@sshleifer)

## Upgrade Command
```bash
python -m pip install 'transformers==3.4.0'
```